Diversity or Disadvantage?
Investigating neighbourhood disorder and ethnic heterogeneity
using the British Crime Survey
Liz Twigg (University of Portsmouth)
Joanna Taylor (University of Portsmouth)
John Mohan (University of Southampton)
ESRC Research Methods Festival, July 6th 2010
Background and
Aims
The influence of ethnic diversity on perceptions
Sampson and Raudenbush’s extensive and detailed body of work on
systematic social observation of disorder; perceptions of disorder and
immigrant concentration. (1999; 2004).
But also recent provocative statements by :Goodhart (2004) – “too diverse” -
Putnam (2007) – increased ethnic
heterogeneity is associated with reduced
levels of trust (even for one’s own race)
and solidarity in the short term.
“hunker down”
“There is evidence that the more diverse an area is in racial terms,
the less likely its residents are to feel that they trust each other.
This is an important argument and it is important that we
examine it”
(David Blunkett, 2004)
Recently Paul Wiles (2009) has questioned whether “[Sampson’s
research] would apply in other countries [such as E&W]”
Perceptions of disorder
Studies of disorderly behaviour have a long history, dating back at least to
Shaw and McKay’s 1942 study of Chicago.
In recent years there has been a “resurgence of interest in neighbourhood
disorder in the social sciences”
(Sampson and Raudenbush, 2004, 319).
“Broken windows” (Wilson and Kelling, 1982)
Sampson (2009, 6) reasons “perceptions of disorder constitute a fundamental
dimension of inequality at the neighbourhood level and beyond”.
Ethnic heterogeneity studies in the UK
UK studies of the effect of heterogeneity on people's perceptions of their
neighbourhood environment have produced equivocal results…
Levels of generalised trust (Pennant, 2005)
Levels of social capital (Letki, 2008)
(both based on the Communities and Local Government’s Citizenship Survey)
Levels of interpersonal trust (Sturgis et al, BJPS forthcoming)
(based on the Dept. Culture, Media and Sport’s Taking Part Survey)
The research: Three measures of disorder
1. Antisocial behaviour
2. Collective efficacy (social cohesion and trust;
informal social control)
3. Perceptions of (local and national) crime trends
Research summarised here
1. Antisocial behaviour
2. Collective efficacy (social cohesion and trust;
informal social control)
3. Perceptions of (local and national) crime trends
Defining collective efficacy
Sampson et al. (1997, p918) defined collective efficacy as
“social cohesion amongst neighbours combined with their
willingness to intervene on behalf of the common good”
and measured it by combining two Likert scales:
(i) Social cohesion and trust
(ii) Informal social control
Aims of the Study
1. We investigate whether, after controlling for individual characteristics
and local socio-economic conditions , we can identify an effect of local
levels of ethnic heterogeneity on perceptions of collective efficacy.
2. In line with concerns expressed in previous literature we explore
whether the effect of deprivation or diversity is more influential.
Aims of the Study
1. We investigate whether, after controlling for individual characteristics
and local socio-economic conditions , we can identify an effect of local
levels of ethnic heterogeneity on perceptions of collective efficacy.
2.
In line with concerns expressed in previous literature we explore
whether the effect of deprivation or diversity is more influential.
3. In contrast with previous work we analyse the dimensions of
collective efficacy separately to determine whether the same
individual, household and area factors influence both
dimensions.
Defining our…
(i) Dataset
(ii) Dependent variables
(iii) Independent variables
The dataset - The British Crime Survey (BCS)
 The BCS is a victimisation survey which asks respondents about their own experiences of crime .
 Primarily designed to capture ‘the dark figure of crime’.
 Also asks many questions on people’s perceptions of their neighbourhood social environment.
 First sweep 1982 – continuous since 2001/02.
 This study based on the 2006/07 sweep with a sample size c47,000 and response rate of 75%.
 Stratified and clustered random sample of adults living in private households.
 Under special license the HO granted us access to a new version of the dataset with Super Output
Areas (SOAs) attached allowing us to attach other data sources such as the Census.
The first dependent variable - social cohesion and trust
The second dependent variable - informal social control
The independent variables
Individual / household level
From the BCS
Gender
Age
Ethnicity
Marital status
Recent victim of BCS crime
Education
Health
Socio-Economic Classification
Household income
Tenure
Accommodation type
Length of time at current address
The independent variables
Individual / household level
From the BCS
Gender
Age
Ethnicity
Marital status
Recent victim of BCS crime
Education
Health
Socio-Economic Classification
Household income
Tenure
Accommodation type
Length of time at current address
Area level
Attached via the SOA codes
Level of urbanisation
Population turnover
Percentage of teenagers
Deprivation
Ethnic heterogeneity
Measure of ethnic heterogeneity
Theil Entropy Score
i stands for a neighbourhood area .
r stands for the following ethnic groups (a) white, (b) mixed, (c) Asian or
Asian British, (d) black or black British, and (e) Chinese or other.
πri represents the proportion of group r in area i (2001 Census).
Modelling strategy
British Crime Survey?
We were given access to a special licence version of
the 2006/7 that had Lower Layer Super Output Area
codes attached to each respondent.
We could therefore link in other datasets:• 2001 census information
• 2007 Indices of Multiple Deprivation
• Cross government rural and urban area
classification
Using Small Area
Identifiers
Problems using LSOAs as a ‘level’
Minimum
population
.
Mean
population
Number in
E&W
Lower layer
1,000
1,500
34,378
Middle layer
5,000
7,200
7,193
LSOAs
•Too sparse clustering; too few
respondents per area
•Don’t necessarily equate to
‘neighbourhood’
MSOAs
•Similar in size to the definition of local
area used in the wording of the ASB
questions;
•Consistent with similar research on the
fear of crime questions;
•Clustering is sufficient for ML models
•Weighted population averages for IMD
data
Multivariate Multilevel Modelling
Advantages of Multivariate Multilevel Modelling
The influence of any one independent variable can be assessed
simultaneously for each dimension of collective efficacy.
Can test whether the effect of any one independent variable is
statistically significant for social cohesion and trust compared with
informal social control.
The multivariate multilevel model estimates higher-level covariance
terms. This joint covariance can illustrate the extent and manner in
which the two dimensions of collective efficacy covary across
geographies (see paper).
Results
Individual and household factors
Social cohesion
and trust
β
SE(β)
Informal social
control
β
SE(β)
Factors which decrease perceived levels of collective efficacy
Male
0.12
0.05
0.39
0.07
Social rented sector
0.67
0.08
0.61
0.10
Living in a flat
0.23
0.09
0.44
0.12
Factors which increase perceived levels of collective efficacy
Age (centred around 50.4)
-0.03
0.00
-0.01
0.00
Asian or Asian British
-0.67
0.16
-0.20
0.20
Moved house within last five years
-0.13
0.06
-0.15
0.08
High income (£40k plus)
-0.49
0.08
-0.52
0.10
Neighbourhood factors
Social cohesion
and trust
β
SE(β)
Informal social
control
β
SE(β)
Factors which decrease perceived levels of collective efficacy
Deprivation
0.36
0.03
0.35
0.04
Ethnic heterogeneity
0.15
0.04
0.22
0.06
Turnover
0.08
0.03
0.04
0.04
Percentage of young people
0.08
0.03
0.06
0.04
Factors which increase perceived levels of collective efficacy
Town or fringe
-0.43
0.10
-0.43
0.13
Village, hamlet or isolated dwelling
-0.91
0.10
-1.30
0.13
Neighbourhood factors
Social cohesion and
trust
β
SE(β)
Exp(β)
Informal social control
β
SE(β)
Exp(β)
Factors which decrease perceived levels of collective efficacy
Deprivation
0.36
0.03
1.43
0.35
0.04
1.42
Ethnic heterogeneity
0.15
0.04
1.16
0.22
0.06
1.24
Turnover
0.08
0.03
1.08
0.04
0.04
1.04
Percentage of young people
0.08
0.03
1.09
0.06
0.04
1.06
Which is most influential – diversity or deprivation?
Factors which increase levels of collective efficacy
Town or fringe
-0.43
0.10
0.65
-0.43
0.13
0.65
Village, hamlet or isolated dwelling
-0.91
0.10
0.40
-1.30
0.13
0.27
Deprivation or diversity?
Statistical significance versus substantive importance
Although the Theil entropy score is statistically significant, it is not substantively important
– it only explains a very small proportion of the MSOA level variation (1% for social
cohesion and trust and 0.3% for informal social control).
The level of deprivation explains substantially more area level variation (19% and 7%
respectively).
Mediating effects
Based on the methodology of Sampson and Raudenbush (1999) we found that ethnic
heterogeneity does not mediate the relationship between collective efficacy and
neighbourhood deprivation.
Restricting analysis to areas with high levels of ethnic heterogeneity
Following the work of Sampson et al (1997, p923) we replicated our model but based
only on the MSOAs with the highest levels of diversity. In these areas deprivation still had
a significant adverse association with collective efficacy.
Conclusions
 Similar individual, household and area factors influence both social cohesion
and trust and informal social control. However, combining them into an overall
measure of collective efficacy masks important differences.
 While both diversity and disadvantage are statistically associated with reduced
levels of social cohesion and trust and informal social control, greater substantive
importance is attached to neighbourhood disadvantage.
Summary of other aspects of the research
ANTISOCIAL BEHAVIOUR
 Different types of people living in the same small area have very different perceptions of
antisocial behaviour.
 Levels of actual crime and disorder
influence perceptions
 Deprivation, rather than diversity,
perceived high levels of ASB
PERCEPTIONS OF LOCAL AND NATIONAL CRIME TRENDS
 Socio-demographic background; newspaper readership perceptions of national crime
trends
 Pessimistic views on localised related to individual crime victimisation.
 No negative effects for ethnic diversity - reduced likelihood of perceiving rising levels of
national crime
Diversity or Disadvantage?
Investigating neighbourhood disorder and
ethnic heterogeneity
using the British Crime Survey
Any questions?